Improve road traffic estimation

[gpetit]

Hi Geoclimate team,

As discussed with @ebocher and @pierromond we would like to improve the way the road traffic is deduced, with OSM data

Based on traffic data extracted from the AVATAR platform we now have a link between road traffic and OSM attributes.
Using this, we have created a regression tree (more or less 5 levels, using less than 20 attributes) that could be used to estimate the traffic on the all french road network.

The question is now to see how we can integrate this tree into Geoclimate and also how to give the possibility to the users to choose the method used to estimate the traffic (ours, for France, or the default one for the rest of the world)

In addition here are some comments or ideas:

• in the output road table, add columns to store:
  • the geometry length
  • the centrality / connectivity / ... indexes
  • the sinuosity index
• in the road_traffic table, keep the OSM_ID column (which is in the road table)

@nicolas-f

[ebocher]

I'm on it

[pierromond]

I did some advances on my side too :) As soon as you get randomForest working for road segments, it will be nice to have a call I guess ;)

Speak soon and thank you again !